Add new xtf.openshift.namespace.per.testcase property which will exec…

[mnovak1]

…ute every test case in its own namespace. Note this property has limitation that consuming test suite must not create static Openshift instances as those are created before name of test case and thus the namespace for the test is known.

Please make sure your PR meets the following requirements:

• Pull Request contains a description of the changes
• Pull Request does not include fixes for multiple issues/topics
• Code is formatted, imports ordered, code compiles and tests are passing
• Code is self-descriptive and/or documented

CI runs (no regression):

-Dxtf.openshift.namespace.per.testcase=false:
https://jenkins.eapqe.psi.redhat.com/job/eap-8.x-openshift-ts-face/251//artifact/output.html

-Dxtf.openshift.namespace.per.testcase=true:
https://jenkins.eapqe.psi.redhat.com/job/eap-8.x-openshift-ts-face/252//artifact/output.html

[mchoma]

@mnovak1 Maybe I would appreciate some introduction how this is implemented. So I better understand it if this can be simplified even more. Overall idea seems OK to me and this seems on good way. But I feel code could be better organized to be even simpler. I drop some random thoughts, but maybe they are not relevant and I am missing something.

[mnovak1]

@mchoma I've addressed all the last comments and added CI runs. Could you review it. please?

[mchoma]

Ok I run through changes and I do not have any signifacnt objection. But truth is I am looking for long time on MR. Maybe it would need some fresh view :).

Just to double check. For someone who is not interested in using xtf.openshift.namespace.per.testcase=true it has no migration consequences?

[mnovak1]

Just to double check. For someone who is not interested in using xtf.openshift.namespace.per.testcase=true it has no migration consequences?

Besides use of the static OpenShift instance and use of NamespaceManager there is no migration issue.

[mchoma]

Approving.

[simkam]

TBH I think that mapping should be done at a different level. TestCase to namespaced OpenShiftClient rather than TestCase to OpenShiftClient to namespace.

But there might be reasons to go this way.

Quite a lot of static fields might also prove to be problematic in the future.

Anyway, approving.

[mnovak1]

Thanks, I'll merge it.

TestCase to namespaced OpenShiftClient rather than TestCase to OpenShiftClient to namespace.

It's more mapping TestCase -> namespace -> OpenShiftClient. I think we've discussed it somewhere already. Motivation is to cache OpenShiftClient as much as possible and making it optimal when this feature is ON or OFF

Mapping TestCase to namespaced OpenShiftClient would work well when this feature is ON. But if it's OFF then it's not desirable to create as many OpenShiftClients as there are TestCases as most likely there would be just 2 OpenShiftClient instances for master and admin.
Mapping TestCase -> namespace -> OpenShiftClient solves this problem as when its' OFF then all TestCases has the same namespace and thus namespace maps again just to 2 OpenShiftClients.